Posterior lenticule laser scan influence the ultra-early postoperative visual acuity and quality of SMILE

Abstract

This is a prospective case-control study investigated the factors influencing ultra-early visual acuity and quality after Small Incision Lenticule Extraction (SMILE) surgery, with a specific focus on the morphology and distribution of "black clefts" observed in posterior lenticule laser scanning images. A total of 174 patients who underwent SMILE surgery were included, and their preoperative ocular clinical parameters, visual quality indicators, and posterior lenticule laser scanning images were analyzed. The morphological features of black clefts were calculated using image processing techniques. Univariate and multivariate logistic analyses were conducted to identify factors impacting ultra-early visual acuity recovery, while correlation analysis was used to explore the association between black cleft morphology and postoperative visual quality. The results showed that although patients experienced significant visual acuity improvement one day after surgery, 26% had poor early recovery. Postoperative visual quality parameters were significantly lower, while the Object Scatter Index (OSI) was higher compared to preoperative values. Patients with poor visual recovery had a higher number of irregular black clefts in their laser scanning images. Logistic regression analysis revealed that intraoperative morphological parameters of black clefts were independent factors influencing early postoperative visual acuity recovery. Additionally, correlation analysis demonstrated a negative correlation between black cleft morphology and postoperative visual quality parameters, and a positive correlation with OSI. In conclusion, the morphology and distribution of black clefts during SMILE surgery significantly affect ultra-early postoperative visual acuity and quality, with increased variability in black cleft area leading to higher OSI, thereby impacting visual quality recovery.

Keywords: Black clefts; Morphology; SMILE; Scanning image; Visual quality.

Fig. 1

Image processing and Clinical outcomes of SIMLE surgery. (A) A streamlined flowchart illustrating the image processing. (B) Comparison between postoperative uncorrected distance visual acuity (UDVA) and preoperative corrected distance visual acuity (CDVA). (C) Variation in preoperative CDVA compared to postoperative UDVA. (D) Relationship between attempted spherical equivalent (SE) refraction change and achieved SE refraction change. E: Distribution of postoperative accuracy in spherical equivalent refraction.

Fig. 2

Example images and Differential analysis. (A) Example of posterior lenticule laser scanning images and corresponding segmentation of black clefts in good visual recovery group. (B) Example of posterior lenticule laser scanning images and corresponding segmentation of black clefts in poor visual recovery group. (C-K) The box plots of num_count (2), area_max (D), area_mean (E), diameter_max (F), diameter_mean (G), perimeter_max (H), perimeter_mean (I), and area_ratio (J) between good and poor visual recovery groups.

Fig. 3

Differential analysis of visual quality. (A) The box plots of visual quality parameters (MTF, SR, IOS, VA100%, VA20%, VA9%) between good and poor visual recovery groups. (B) The differential analysis of delta visual quality parameters (MTF, SR, IOS, VA100%, VA20%, VA9%) between good and poor visual recovery groups. ** means P < 0.01; ns mean not significant.

Fig. 4

Logistic Analysis. (A) The forest plot of univariate logistic analysis. (B) The forest plot of multivariate logistic analysis. * means P < 0.05; ** means P < 0.01.

Fig. 5

Correlation analysis and Construction of regression model. (A) The heatmap of correlation for morphological features with visual quality parameters. (B) The heatmap of correlation for morphological features with delta visual quality parameters. (C) The values of Root Mean Squared Error (RMSE) for 9 machine-learning methods. (D) The values of Mean Absolute Error (MAE) for 9 machine-learning methods. (E) The values of R Squared for 9 machine-learning methods. (F) The architecture of the Multi-Layer Perceptron (MLP) model (G) The distribution of Shap values for 9 machine-learning methods. (H) The correlation between predict and actual OSI in training dataset. (I) The correlation between predict and actual OSI in testing dataset. (J) The correlation between predict and actual OSI in all dataset.